China Research Report Sentiment (Sample)

Datayes cooperates with more than 98% brokers in China, including key institutes, and 98% analysts. There are over 2.75 million research reports dating back to Jan. 2006 with timely daily updates, covering more than 5300 China A-shares listed companies.

The research report sentiment data utilizes Datayes’ self-developed technologies (OCR, viewpoint extraction, NLP sentiment analysis, etc.) to extract multiple core views from each research report and calculate the sentiment score for each viewpoint using a deep learning model. The sentiment data contains two tables.

Sample Table:
Total research report sentiment score table: describes the basic information of the company's research report, such as title, broker, analyst, stock code, rating, etc., and the total sentiment score of the research report.

Research report core view sentiment table: describes multiple core view sentiment-related data of each research report, specifically containing detailed information such as sentiment classification, general positive sentiment probability, substantial positive sentiment probability, negative sentiment probability, neutral sentiment probability, and sentiment score of each view (identified by record ID).

Sample Field for Research report core view sentiment table
1 ID
2 RESEARCH_REPORT_ID
3 RECORD_ID

4 SOURCE

5 RPT_INSERT_TIME

6 SENTIMENT_CLASSIFY

7 HALF_POSITIVE_PROBABILITY

8 POSITIVE_PROBABILITY

9 NEGATIVE_PROBABILITY

10 NEUTRAL_PROBABILITY

11 SENTIMENT_SCORE

12 UPDATE_TIME

搜集汇总

背景与挑战

背景概述

该数据集涵盖2006年以来5300余家中国A股上市公司的275万份券商研报，通过OCR、NLP等技术提取核心观点并计算情感得分，包含研报基础信息表和核心观点情感分析表，后者记录情感分类、正负向概率及情感评分等详细维度。